Amanai Changelog

Pricing Engine v2

Credits now go about 2× further for typical usage, with consistent component-based rates.

Historical release record. The tables below describe Pricing Engine v2 at launch. For current model-specific multipliers, use Billing & pricing or the dashboard calculator.

Billing

Pricing Engine v2

Credits now go about 2× further for typical usage, with consistent component-based rates and no migration required. Actual consumption still depends on the model, output volume, cache behavior, and request size.

Overview

Pricing Engine v2 makes credit consumption more consistent across Amanai's model catalog. Input, cache reads, and generated output are calculated separately, while the same public API keys and request formats continue to work.

Typical active usage lasts about twice as long in historical replay.
Existing balances keep the same displayed amount and expiration date.
No API key, model ID, or client configuration changes are required.
Ultra adds a higher-volume option with 1 billion credits.

Pricing v1 vs Pricing v2

The table summarizes the product-level changes. “About 2×” is an observed typical-use estimate, not a guaranteed rate for every request or model.

AreaPricing v1Pricing v2
Typical credit lifespanBaselineAbout 2× longer for typical usage
Output rateVaried by modelStandardized at 5× the model input rate
Cache-read rateVaried by modelStandardized at 15% of the model input rate
Minimum chargeNo request minimum1,000 credits per billable request
Existing balancesUsed Pricing v1Same balance, automatically uses Pricing v2
Available packsStarter, Pro, MaxStarter, Pro, Max, Ultra
Interrupted deliveryErrors generally refundedNo delivery is refunded; delivered partial usage is settled

Model examples

These examples show m_in / m_out / m_cache. Values are rounded to two decimal places for display; billing retains full internal precision.

ModelPricing v1Pricing v2
m_inm_outm_cachem_inm_outm_cache
amanai/glm-5.2 13.243.50.49 6.0330.170.3
amanai/glm-5.1 9.2430.450.34 4.2321.140.21
amanai/glm-5.0 6.621.750.25 3.0215.10.15
amanai/glm-5.0-turbo 7.9226.10.29 3.6218.120.18
amanai/glm-4.7 5.2817.40.2 2.4212.080.12
amanai/glm-5v-turbo 9.2430.450.34 4.2321.140.21
amanai/kimi-k2.7 10.5634.80.39 4.8324.160.24
amanai/kimi-k2.6 9.531.320.35 4.3521.730.22
amanai/kimi-k2.5 8.528.040.31 3.8919.440.19
amanai/deepseek-v4-pro 6.613.90.25 3.0215.10.15
amanai/deepseek-v4-flash 2.14.40.08 0.964.80.05
amanai/minimax-m3 4.6215.230.17 2.1110.570.11
amanai/minimax-m2.7 3.9613.050.15 1.819.060.09
amanai/hy3-preview 7.2623.930.27 3.3216.610.17
amanai/grok-4.5 12.5441.330.47 5.7428.690.29
amanai/grok-4.20 8260.5 3.6618.30.18
amanai/grok-4.3-low 9290.5 4.1220.590.21
amanai/grok-4.3-medium 10330.52 4.5822.880.23
amanai/grok-build 9.5310.52 4.3521.730.22
amanai/grok-4.20-console 9290.52 4.1220.590.21
amanai/grok-4.20-reasoning 10.5340.52 4.824.020.24
amanai/grok-multi-console 11360.52 5.0325.160.25
amanai/grok-multi-medium 12390.54 5.4927.450.27
amanai/grok-multi-high 13420.55 5.9529.740.3
amanai/grok-multi-xhigh 14450.55 6.4132.030.32
amanai/gpt-5.4-mini 1.424.690.05 0.653.250.03
amanai/gpt-5.4 1.565.160.06 0.713.570.04
amanai/gpt-5.5 2.849.340.11 3150.15
amanai/gpt-5.6-sol 13.8645.680.51 6.3431.70.32
amanai/gpt-5.6-terra 13.0643.020.48 5.9729.870.3
amanai/gpt-5.6-luna 13.4644.350.5 6.1630.790.31
amanai/claude-opus-4.7 39.6130.51.47 22.65113.231.13
amanai/claude-opus-4.6 39.6130.51.47 22.65113.231.13
amanai/claude-sonnet-4.6 26.4870.98 15.175.490.75
amanai/claude-haiku-4.5 12400.5 6.8634.310.34
amanai/grok-imagine-liteImage · flat 100,000 credits/imageImage · flat 100,000 credits/image

The table is a historical Pricing v1 snapshot compared with the current Pricing v2 catalog. See Billing & pricing for the current public explanation.

Existing balances

Credits already held by an account were not converted, removed, or moved into a separate wallet. The same balance now pays Pricing v2 rates for requests that start after the release.

Ultra

Ultra joins Starter, Pro, and Max with 1,000,000,000 credits for Rp150,000, valid for 30 days. It includes access to all models, 320 RPM, and up to 24 concurrent requests.

Billing behavior

Billable text requests have a 1,000-credit minimum.
Cache creation remains part of normal input; cache reads receive the discounted cache rate.
The gateway reserves a worst-case estimate and refunds unused reserve after settlement.
Failures before any delivery are fully refunded.
Partial output already delivered is charged from measured or deterministically estimated usage.
Reasoning tokens are not charged as a separate component.

Notes

Actual savings vary. Output-heavy requests, model choice, cache-hit rate, and the 1,000-credit minimum all affect final consumption. Very small requests can cost more than under Pricing v1 because of the minimum charge.

For the complete formula and current model rates, see Billing & pricing.

← Membership & credits policy